-
oa Machine Learning-Informed Reservoir Simulation for Performance Mapping of CO2-EOR and Storage in the SACROC Field
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, World CCUS Conference 2025, Sep 2025, Volume 2025, p.1 - 5
Abstract
This study presents an innovative approach to integrating machine learning (ML) with reservoir simulation to enhance the predictive accuracy and efficiency of CO2-enhanced oil recovery (CO2-EOR) and storage in the SACROC field of the Permian Basin. Traditional reservoir simulation methods are computationally expensive and time-consuming, particularly when dealing with complex geological formations. By leveraging ML algorithms—Random Forest, Gradient Boosting, and XGBoost—this study aims to provide faster and more accurate predictions of oil and CO2 saturation distributions. A dataset of 13,600 data points, including key reservoir parameters such as permeability and porosity, is used to train the models. Performance evaluation through statistical metrics such as RMSE, MAE, and R2 identifies Random Forest as the most effective model, achieving the lowest prediction error and highest accuracy. The results demonstrate the capability of ML models to accurately map reservoir dynamics over short- and long-term periods, offering valuable insights for optimizing CO2 injection strategies and improving carbon capture, utilization, and storage (CCUS) efficiency. The findings highlight the potential of ML-driven frameworks to support decision-making processes in subsurface engineering, ultimately contributing to sustainable energy development and climate change mitigation efforts.
